Technology and Privacy:
The New Landscape

edited by
Philip E. Agre
University of California, San Diego

Marc Rotenberg
Electronic Privacy Information Center

MIT Press, 1997

Privacy is the capacity to negotiate social relationships by controlling access to personal information. As laws, policies, and technological design increasingly structure people's relationships with social institutions, individual privacy faces new threats and new opportunities. Over the last several years, the realm of technology and privacy has been transformed, creating a landscape that is both dangerous and encouraging. Significant changes include large increases in communications bandwidths; the widespread adoption of computer networking and public-key cryptography; mathematical innovations that promise a vast family of protocols for protecting identity in complex transactions; new digital media that support a wide range of social relationships; a new generation of technologically sophisticated privacy activists; a massive body of practical experience in the development and application of data-protection laws; and the rapid globalization of manufacturing, culture, and policy making.

The essays in this book provide a new conceptual framework for the analysis and debate of privacy policy and for the design and development of information systems. The authors are international experts in the technical, economic, and political aspects of privacy; the book's strength is its synthesis of the three. The book provides equally strong analyses of privacy issues in the United States, Canada, and Europe.


Philip E. Agre, Beyond the mirror world: Privacy and the representational practices of computing

Victoria Bellotti, Design for privacy in multimedia computing and communications environments

Colin J. Bennett, Convergence revisited: Towards a global policy for personal data protection

Herbert Burkert, Privacy enhancing technologies: Typology, vision, critique

Simon G. Davies, Re-engineering the privacy right: How privacy has been transformed from a right to a commodity

David H. Flaherty, Controlling surveillance: Can privacy protection be made effective?

Robert Gellman, Does privacy law work?

Viktor Mayer-Schoenberger, Generational development of data protection in Europe

David J. Phillips, Cryptography, secrets, and the structuring of trust

Rohan Samarajiva, Interactivity as though privacy mattered

Introduction (8800 words)

(Please do not quote from this version, which changed slightly in proof.)

1. Introduction

Our premise in organizing this volume is that, since the 1980s, the policy debate around technology and privacy has been transformed. Tectonic shifts in the technical, economic, and policy domains have brought us to a new landscape that is more variegated, more dangerous, and more hopeful than before. These shifts include the emergence of digital communications networks on a global scale; emerging technologies for protecting communications and personal identity; new digital media that support a wide range of social relationships; a generation of technologically sophisticated privacy activists; a growing body of practical experience in developing and applying data protection laws; and the rapid globalization of manufacturing, culture, and the policy process. The goal of this volume is to describe this emerging landscape. By bringing together perspectives from political science, law, sociology, communications, and human-computer interaction, we hope to offer conceptual frameworks whose usefulness may outlive the frenetically changing details of particular cases. We believe that in the years ahead the public will increasingly confront important choices about law, technology, and institutional practice. This volume offers a starting point for analysis of these choices.

The purpose of this introduction is to summarize and synthesize the picture of this new landscape that the contributors have drawn. First, however, I should make clear what we have not done. We have not attempted to replace the foundational analysis of privacy that has already been admirably undertaken by Allen (1988), Schoeman (1984), and Westin (1967). We have not replicated the fine investigative work of Burnham (1983) and Smith (1979). Nor, unlike Lyon and Zureik (1996), have we tried to place the issues of privacy and surveillance in their broadest sociological context. Our work is organized conceptually and not by area of concern (medical, financial, marketing, workplace, political repression, and so on). Although our case studies are drawn from several countries, our method is not systematically comparative (see Bennett 1992, Flaherty 1989, and Nugter 1990). We have not attempted a complete survey of the issues that fall in the broad intersection of "technology" and "privacy." By "technology," for example, we mean information and communications technology; we do not address the concerns raised by biological technologies such as genetic analysis (Gostin 1995). Our concern with the interactions among technology, economics, and policy complements Smith's (1994) study of organizational issues and Regan's (1995) more detailed analysis of the legislative process. Nor, finally, do we provide a general theory of privacy or detailed policy proposals. We hope that our work will be helpful in framing the new policy debate, and we have analyzed several aspects of the development of privacy policy to date.

2. The New Landscape

Mayer-Schoenberger's chapter describes the configuration of technology and privacy issues in the late 1960s and the early 1970s. In that period, privacy concerns focused on a small number of large centralized databases; although instrumental to the construction of the modern welfare state, these databases also recalled the role of centralized files in the fascist era. In the United States, concern about privacy arose through popular works by Ernst and Schwartz (1962), Brenton (1964), and Packard (1964), as well as a detailed scholarly treatment by Westin (1967). In each case, though, the general form of the response was the same -- an enforceable code of practice that came to be known as data protection in Europe and privacy protection in the United States. The premise underlying the Code of Fair Information Practices was the same in both places: organizations that collected personal information about individuals had certain responsibilities, and individuals had rights against organizations in possession of personal information. In some instances, these practices were codified by professions or industry associations. In other instances they were reduced to law. As a general matter, the focus was the centralized collection of data, specified in place and time, and under the specific responsibility of a known individual or organization. (These principles and their implementation are described by Gellman. I will use the term "data protection" here.) Data protection does not seek to influence the basic architecture of computer systems. Instead, it abstracts from that architecture to specify a series of policies about the creation, handling, and disposition of personal data. Mayer-Schoenberger, Bennett, and Flaherty describe the subsequent evolution of the data protection model. This model is by no means obsolete, but the world to which it originally responded has changed enormously.

Some of these changes are technical. Databases of personal information have grown exponentially in number and in variety. The techniques for constructing these databases have not changed in any fundamental way, but the techniques for using them have multiplied. Data-mining algorithms, for example, can extract commercially meaningful patterns from extremely large amounts of information. Market-segmentation methods permit organizations to target their attention to precisely defined subgroups (Gandy 1993). Contests, mass mailings, and other promotions are routinely organized for the sole purpose of gathering lists of individuals with defined interests. More data is gathered surreptitiously from individuals or sold by third parties.

The pervasive spread of computer networking has had numerous effects. It is now easier to merge databases. As Bennett observes, personal information now routinely flows across jurisdictional boundaries. Computer networking also provides an infrastructure for a wide variety of technologies that track the movements of people and things (Agre 1994). Many of these technologies depend on digital wireless communications and advanced sensors. Intelligent Transportation Systems, for example, presuppose the capacity to monitor traffic patterns across a broad geographic area (Branscomb and Keller 1996). These systems also exemplify the spread of databases whose contents maintain a real-time correspondence to the real-world circumstances that they represent. These computerized mediations of personal identity have become so extensive that some authors speak of the emergence of a "digital persona" that is integral to the construction of the social individual (Clarke 1994).

Computer networking also provides the basis for a new generation of advanced communications media. In the context of the analog telephone system, privacy concerns (e.g., wiretapping and the abuse of records of subscribers' calls) were largely circumscribed by the system's architecture. Newer media, such as the Internet and the online services discussed by Samarajiva, capture more detailed information about their users in digital form. Moreover, the media spaces that Bellotti describes are woven into their users' lives in a more intimate way than older communications technologies. Digital technology also increases both the capacity of law-enforcement authorities to monitor communications and the capacity of subscribers to protect them.

At the same time, the new media have provided the technical foundation for a new public sphere. Privacy activists and concerned technologists have used the Internet to organize themselves, broadcast information, and circulate software instantaneously without regard to jurisdictional boundaries. Low-cost electronic-mail alerts have been used in campaigns against consumer databases, expanded wiretapping capabilities, and government initiatives to regulate access to strong cryptography. Public-policy issues that would previously have been confined to a small community of specialists are now contested by tens of thousands of individuals. Although the success of these tactics in affecting policy decisions has not yet been evaluated, the trend toward greater public involvement has given the technology a powerful symbolic value.

Potentially the most significant technical innovation, though, is a class of privacy-enhancing technologies (PETs). Beginning with the publication of the first public-key cryptographic methods in the 1970s, mathematicians have constructed a formidable array of protocols for communicating and conducting transactions while controlling access to sensitive information. These techniques have become practical enough to be used in mass-market products, and Phillips analyzes some of the sharp conflicts that have been provoked by attempts to propagate them. PETs also mark a significant philosophical shift. By applying advanced mathematics to the protection of privacy, they disrupt the conventional pessimistic association between technology and social control. No longer are privacy advocates in the position of resisting technology as such, and no longer (as Burkert observes) can objectives of social control (if there are any) be hidden beneath the mask of technical necessity. As a result, policy debates have been opened where many had assumed that none would exist, and the simple trade-off between privacy and functionality has given way to a more complex trade-off among potentially numerous combinations of architecture and policy choices.

Other significant changes are political and economic. The data protection model has matured. Privacy commissioners such as Flaherty have rendered hundreds of decisions in particular cases, and the nature and the limitations of the privacy commissioner's role have been clarified. It has become possible to ask how the effectiveness of privacy policies might be evaluated, although (as both Flaherty and Bennett observe) few useful methods have emerged for doing so. Pressures have arisen to tailor data protection laws to the myriad circumstances in which they are applied, with the result that sectoral regulation has spread. In the United States, as Gellman observes, the sectoral approach has been the norm by default, with little uniformity in regulatory conception or method across the various industries. In most other industrial countries, by contrast, sectoral regulation has arisen through the adaptation and tailoring of a uniform regulatory philosophy.

This contrast reflects another, deeper divide. Bennett describes the powerful forces working toward a global convergence of the conceptual content and the legal instruments of privacy policy. These forces include commonalities of technology, a well-networked global policy community, and the strictures on cross-border flows of personal data in the European Union's Data Protection Directive. While the United States has moved slowly to establish formal privacy mechanisms and standardize privacy practices over the last two decades, it now appears that the globalization of markets, the growing pervasiveness of the Internet, and the implementation of the Data Protection Directive will bring new pressures to bear on the American privacy regime.

The evolution of privacy policy, meanwhile, has interacted with individual nations' political philosophies. Mayer-Schoenberger argues that this interaction should be viewed not on a nation-by-nation basis but rather as the expression of a series of partial accommodations between the uniform regulation of data handling and liberal political values that tend to define privacy issues in terms of localized interactions among individuals. (This tension runs throughout the contemporary debate and will recur in various guises throughout this introduction.)

One constant across this history is the notorious difficulty of defining the concept of privacy. The lack of satisfactory definitions has obstructed public debate by making it hard to support detailed policy prescriptions with logical arguments from accepted moral premises. Attempts to ground privacy rights in first principles have foundered, suggesting their inherent complexity as social goods. Bennett points out that privacy is more difficult to measure than other objects of public concern, such as environmental pollution. The extreme lack of transparency in societal transfers of personal data, moreover, gives the issue a nebulous character. Citizens may be aware that they suffer harm from the circulation of computerized information about them, but they usually cannot reconstruct the connections between cause and effect. This may account in part for the striking mismatch between public expression of concern in opinion polls and the almost complete absence of popular mobilization in support of privacy rights.

One result of this unsatisfactory situation is that the debate has often returned to the basics. Mayer-Schoenberger and Davies both remark on the gap between the technical concept of data protection and the legal and moral concept of privacy, but they assign different significance to it. For Mayer-Schoenberger, the concept of data protection is well fitted to the values of the welfare state. Davies, however, focuses on the range of issues that data protection appears to leave out, and he regards the narrowly technical discourse of data protection as ill suited to the robust popular debate that the issues deserve.

The basic picture, then, is as follows: Privacy issues have begun to arise in more various and more intimate ways, a greater range of design and policy options are available, and some decisions must therefore be made that are both fundamental and extraordinarily complicated. Perhaps the most basic of these strategic decisions concerns the direction of technical means for protecting privacy. One approach, exemplified by Bellotti's study, is to provide individuals with a range of means by which to control access to information about themselves. Another approach, discussed in detail by Phillips and by Burkert, is to prevent the abuse of personal information from the start through the application of privacy-enhancing technologies that prevent sensitive data from being personally identifiable. PET's may also be appealing because, unlike privacy codes, they are to some extent self-enforcing. Although PET's may not eliminate the need for ongoing regulatory intervention, they seem likely to reduce it.

3. Negotiated Relationships

Ideas about privacy have often been challenged by new technologies. The existing ideas arose as culturally specific ways of articulating the interests that have been wronged in particular situations. As a result, cultural ideas about privacy will always tacitly presuppose a certain social and technological environment -- an environment in which those kinds of wrongs can occur and in which other kinds of wrongs are either impossible, impractical, or incapable of yielding any benefit to a wrongdoer. As new technologies are adopted and incorporated into the routines of daily life, new wrongs can occur, and these wrongs are often found to invalidate the tacit presuppositions on which ideas about privacy had formerly been based. The development of law illustrates this. In a landmark 1967 case, Katz v United States (389 US 347), the US Supreme Court found that a warrantless police recording device attached to the outside of a telephone booth violated the Fourth Amendment's protection against unreasonable searches and seizures. This protection had formerly been construed primarily in cases involving intrusion into a physical place, but the justices famously held that the Fourth Amendment "protects people, not places" (at 351).

The moral interest at stake in data protection regulation has seemed unclear to many. Turkington (1990), among others, has suggested identifying this interest as "informational privacy." Another complementary approach can be understood by returning to Clarke's (1994) notion of the "digital persona" that has increasingly become part of an individual's social identity. From this perspective, control over personal information is control over an aspect of the identity one projects to the world, and the right to privacy is the freedom from unreasonable constraints on the construction of one's own identity. This idea is appealing for several reasons: it goes well beyond the static conception of privacy as a right to seclusion or secrecy, it explains why people wish to control personal information, and it promises detailed guidance about what kinds of control they might wish to have.

Bellotti and Samarajiva develop this line of thinking further. Their point of departure is the sociologist Erving Goffman's finely detailed description of the methods by which people project their personae. Goffman (1957) argued that personal identity is not a static collection of attributes but a dynamic, relational process. People construct their identities, he suggested, through a negotiation of boundaries in which the parties reveal personal information selectively according to a tacit moral code that he called the "right and duty of partial display." Goffman developed this theory in settings (e.g., public places) where participants could see one another face-to-face, but it has obvious implications for technology-mediated interactions. In particular, to the extent that a technology shapes individuals' abilities to negotiate their identities, Goffman's theories have implications for that technology's design.

Bellotti and Samarajiva attempt to draw out these implications in different contexts, Bellotti in her experiments with media spaces that interconnect users' workspaces with video and data links and Samarajiva in his study of an online platform being developed in Quebec. These authors explore the conditions under which individuals can exert control and receive feedback over the release of personal information, Bellotti emphasizing technical conditions and Samarajiva emphasizing institutional conditions. Both authors describe conditions under which pathological relationships might arise. Goffman's theory, by helping articulate the nature of the wrongs in such cases, also helps specify how technology might help us avoid them. Technology cannot, of course, guarantee fairness in human relationships, but it can create conditions under which fairness is at least possible; it can also undermine the conditions of fairness.

CNID (also known, somewhat misleadingly, as Caller ID) illustrates the point. CNID is a mechanism by which a switching system can transmit a caller's telephone number to the telephone being called. The recipient's telephone might display the number or use it to index a database. CNID seems to raise conflicting privacy interests: the caller's right to avoid disclosing personal information (her telephone number) and the recipient's right to reduce unwanted intrusions by declining to answer calls from certain numbers. To reconcile these interests, CNID systems generally come equipped with "blocking" options. Callers who do not wish to transmit their telephone number can block CNID; recipients may decline to answer calls that are not accompanied by CNID information. In effect, the negotation of personal identity that once began with conventional telephone greetings ("Hello?"; "Hello, this is Carey; is Antonia there?"; "Oh, hi Carey, it's Antonia; how are you?") now begins with the decision whether to provide CNID information and the decision whether to answer the phone. Sharp conflict, however, arises in regard to the details of the blocking interface. If CNID is blocked by default then most subscribers may never turn it on, thus lessening the value of CNID capture systems to marketing organizations; if CNID is unblocked by default and the blocking option is inconvenient or little-known, callers' privacy may not be adequately protected. In 1995 these considerations motivated a remarkable campaign to inform telephone subscribers in California of their CNID options. The themes of default settings and user education recur in Bellotti's study, in which one system failed to achieve a critical mass of users because so many users had unintentionally kept it turned off.

It is useful to compare this approach to privacy with the approach taken by many proponents of PETs. Representative studies from this volume might be arranged in two dimensions, as shown in figure 1. The horizontal axis distinguishes cases according to the structure of the interaction -- between individual subjects or between a subject and an institution such as a bank. This distinction is obviously artificial, since it excludes intermediate cases, additional relevant parties, and other elements of context. The vertical axis pertains to the conceptualization of privacy, as the negotiation of personal boundaries or as the regulated and conditional access of some authority to sensitive personal information. (A third axis might contrast the normative standpoint of the designer or policymaker -- see Bellotti and Burkert -- with the empirical standpoint of the sociologist -- see Phillips and Samarajiva.) In Phillips's study of public contests over cryptography, the relevant authority is the government and the sensitive information is the cryptographic keys that can reveal the content of personal communications; in Burkert's analysis of technologies for conducting anonymous transactions, the relevant authorities are numerous and the sensitive information is personal identity. These authors' approach is labeled "binary access" because the issue at stake is whether the authority can gain access to the information; this access might be conditional (for example, upon a court order), but once granted it opens up a whole realm of personal information, often covertly, to a broad range of unwelcome uses.

The two rows in figure 1 contrast in instructive ways. Personal boundary negotiation emphasizes personal choice, reciprocity, and fine-grained control over information in the construction of personal identity. Binary access emphasizes individual rights, the defense of interests, and coarse-grained control over information in the protection of personal identity. This distinction partly reflects real differences in the social relationships that two rows describe, but for some authors it also reflects different underlying approaches to privacy problems. Burkert points out that PETs are not inherently confined to the maintenance of anonymity; they could be used as tools for the finer-grained negotiation of identity. Mayer-Schoenberger associates technologies for local choice with erosion of the values of social democracy and emphasizes that individuals have found themselves confused and overwhelmed by a proliferation of mechanisms that require them to make choices about esoteric matters that are more suited to technical analysis and legislation. Other authors, undaunted by this potential, have embraced technologies for local choice as expressions of market freedom. In each case, the actual conditions of fairness and the effective protection of interests in concrete situations are largely unknown.

The new technologies also have implications for conceptions of relationship, trust, and public space. Samarjiva observes that technology and codes of practice determine whether data-based "relationships" between organizations and individuals are fair, or whether they provoke anxiety. These concerns are a traditional motivation for data protection regulation, but they are amplified by technologies that permit organizations to maintain highly customized "relationships" by projecting different organizational personae to different individuals. Such "relationships" easily become asymmetric, with the organization having the greater power to control what information about itself is released while simultaneously obscuring the nature and scope of the information it has obtained about individuals. Phillips describes a controversy over the conditions under which individuals can establish private zones that restrict access by outsiders. A secure telephone line is arguably a precondition for the establishment of an intimate relationship, an interest which has long been regarded as a defining feature of human dignity (e.g., Fried 1968). This concern with the boundaries that are established around a relationship complements Bellotti's concern with the boundaries that are negotiated within a relationship. It also draws attention to the contested nature of those boundaries.

Beneficial relationships are generally held to require trust. As the information infrastructure supports relationships in more complex ways, it also creates the conditions for the construction of trust. Trust has an obvious moral significance, and it is economically significant when sustained business relationships cannot be reduced to periodic zero-sum exchange or specified in advance by contract. Phillips and Burkert emphasize the connection between trust and uncertainty, but they evaluate it differently. For Phillips, trust and uncertainty are complementary; cryptography establishes the boundaries of trust by keeping secrets. Burkert, however, is concerned that this approach reduces trustworthiness to simple reliability, thereby introducing tacit norms against trusting behavior. Just as technology provides the conditions for negotiating relationships, it also provides the conditions for creating trust. Samarajiva points to the institutional conditions by which a technical architecture comes to support these conditions or else evolves toward a regime of coercive surveillance.

Public spaces have traditionally been understood as primary sites for the conduct of politically significant activities, and systematic surveillance of those spaces may threaten the freedom of association. Davies describes the shifting ways in which public discourse in the United Kingdom since the 1980's has constructed the issue of surveillance in public spaces. The introduction of inexpensive video cameras has brought a tension between privacy and personal security, generally to the detriment of the political value of privacy. Bellotti points out that the new technologies are being used in ways that erode the distinction between public and private space and in ways that problematize the very idea of private space by establishing long-lived interconnections among formerly separate spaces. Although occasioned by technological innovations, these observations converge with recent attempts to renegotiate the concepts of public and private and the conceptions of intimate relationship and political discourse that have traditionally gone with them.

Taken together, these considerations describe a complex new terrain around the central issues of voluntariness and coercion. Davies observes that organizations often identify disclosures of personal information as voluntary, even when the consequences of disclosure are unclear, when alternative courses of action are unavailable, or when failures to disclose are accompanied by unreasonable costs. This contested terrain was first mapped by the data protection model, with its concepts of notification and transparency. The purpose of these concepts was to enable individuals to bargain more effectively over the disclosure and use of personal information. The emerging technological options, however, create a more complicated range of alternatives for policy and design. A fundamental division of labor emerges: some decisions about privacy are made collectively at the level of system architecture, and others are made individually through local bargaining. System architecture necessarily embodies social choices about privacy; these choices can make abuses more difficult, but they can also prevent individuals from tailoring their technology-mediated relationships to their particular needs. System architecture, however, rarely suffices to shape social outcomes, and architectural choices must always be complemented by policy measures such as regulations and sectoral codes. The data protection model analyzed the relationship between architecture and policy in a simple, powerful way. Now, however, other possible analyses are becoming perceptible on the horizon.

4. Economic and Technical Scenarios

Privacy issues, then, pertain to the mechanisms through which people define themselves and conduct their relationships with one another. These mechanisms comprise technologies, customs, and laws, and their reliability and fairness are necessary conditions of a just social order. Those who raise concerns about privacy propose, in effect, to challenge the workings of institutions. Disputes about privacy are, among other things, contests to influence the historical evolution of these institutions. Before considering these contests in detail, though, let us consider the economic and technical logics that, some have held, are truly driving the changes that are now under way. My purpose is not to present these scenarios as adequate theories of the phenomena, but rather to make them available for critical examination.

A useful point of departure is Casson's (1994) analysis of the role of business information in the evolution of social institutions. Casson observes that information and markets have a reciprocal relationship: perfect markets require perfect information, but information is usually not free. To the contrary, information is one more good that is traded in the market. As information becomes cheaper (for example through new technology) transaction costs are reduced accordingly and the conditions for perfect markets are better approximated (Coase 1937, Williamson 1975). This theory makes the remarkable prediction that many social institutions, having originated to economize on information costs, will break down, to be replaced by institutions that more closely resemble the market ideal of individually negotiated transactions. Advance ticket sales through networked booking systems, to take a simple example, now supplement box-office queues as a mechanism for allocating movie tickets. On a larger scale, the gradual breakdown of fixed social roles and of the customary responsibilities that go with them is facilitated by technological changes, particularly in communications and in record keeping, that make it easier to establish and evaluate individual reputations.

This theory suggests two consequences for privacy. The first is that individualized transactions must be monitored more closely than custom-bound transactions. By way of illustration, Casson offers a science fiction story about the future of road transportation in a world of very low information costs. Nowadays most roads are provided collectively, and their use is governed by customary mechanisms (such as traffic lights and right-of-way rules) that permit drivers to move toward their destinations without colliding very often. From an economic standpoint, this scheme has obvious inefficiencies: road utilization is uneven, congestion is common, and drivers spend much time waiting. With lower information costs, however, roads could operate more like railroads. Drivers wishing to go from point A to point B would call up a reservation system and bid for available itineraries, each precisely specifying the places and times of driving. Drivers' movements would be still regulated by traffic signals, but the purpose of the signals would now be to keep the drivers within the space-time bounds of the journey they had purchased. Lower-paying drivers would be assigned slower and less scenic routes, other things being equal, than higher-paying drivers. Overall efficiency would be maximized by the market mechanisms embodied in the reservation system. Issues of technical workability aside, Casson points out that such a scheme would raise concerns through its reliance on detailed monitoring of drivers' movements. Nor are these concerns entirely hypothetical, in view of the automatic toll-collection technologies now being deployed as part of the Intelligent Transportation Systems program mentioned above (Agre 1995). Decreasing information costs make toll roads cheaper to operate, thus contributing to their spread. Information costs alone, of course, do not explain the full political and institutional dynamics of these developments, but they do lower one barrier to them.

The second consequence of Casson's theory relates to the evolution of privacy regulation itself. Data protection regulation, on this theory, is essentially a set of customary responsibilities imposed on organizations that gather personal information. The theory further suggests that these responsibilities are economically inefficient, inasmuch as they they preclude individualized negotiations between organizations and individuals about the handling of each individual's information. The "one size fits all" regulations are efficient, however, if the costs of individualized negotiation are high. Data protection regulation is thus similar to the phenomenon of standardized contracts, which also economize on transaction costs, and regulation (as opposed to market competition between different sets of standardized contract terms) is required because the parties to a standardized contract hold asymmetrically incomplete information about the real costs and benefits of various information-handling policies. New information technologies reopen these questions in two ways: by making it economically feasible in some cases for organizations to negotiate the handling of customers' information on a more individualized basis, and by providing policymakers with a broader range of possible information-handling policies to impose on organizations that transact business with the public. Casson's analysis focuses on the first of these points, predicting a transition from generalized regulation to localized negotiation of privacy matters.

These two implications of Casson's theory may seem contradictory: more individualized market transactions require greater monitoring and therefore less privacy, and decreases in transaction costs permit a transition toward local negotiation and thus more efficient allocation of rights to personal information. Both of these contradictory movements are found in reality, however, and both movements are likely to continue. An economic optimist would suggest that the necessary outcome is a high level of potential monitoring whose actual level is regulated by the same allocative mechanisms that regulate everything else in the market. This scenario, however, makes numerous assumptions. It assumes, for example, that market forces will cause economic institutions to become perfectly transparent. This seems unlikely. After all, as Casson points out, information about information is inherently expensive because it is hard to evaluate the quality of information without consuming it. Economic efficiency can even be reduced by advertising and public-relations practices that frustrate consumers' attempts to distinguish between valid and bogus reputations. (Davies's chapter bears on this point in reference to the public construction of privacy issues by interested parties.) Most important, the optimistic economic scenario applies only in the longest possible term, after all transaction costs have been reduced to zero and the whole technical infrastructure of society has been revised to implement the efficient economic regime that results. It says little about the fairness of the constantly shifting and inevitably contested series of institutional arrangements that will govern social life between now and then.

Even if it is accepted, Casson's theory predicts only incremental shifts and should not be interpreted as arguing for a wholesale abandonment of regulation. Indeed, to the extent that transaction costs remain high, the theory argues that traditional protections should be retained. Advances in technology do not necessarily reduce transaction costs, and many such costs may have no relationship to technology. In some cases, as Rotenberg (1996) has pointed out, reductions in transaction costs should actually cause traditional protections to be strengthened. Posner (1981: 256, 264), for example, argues that high transaction costs should relieve magazine publishers and the Bureau of the Census of the obligation to obtain individuals' permission before making certain secondary uses of information about them; as technology lowers transaction costs, this argument is steadily weakened.

Assuming, however, that something approximating the economic scenario comes true, how might the necessary technologies of localized negotiation be implemented? Data protection regulation, as we have seen, was originally motivated by fears about a single centralized government database, and it was subsequently forced to adjust its imagination to accommodate a world of wildly fragmented databases in both the public and the private sector. Immense incentives exist to merge these various databases, however, and many have predicted that the spread of computer networking will finally bring about the original dystopian vision. This vision has not come about on a large scale because of the great difficulty of maintaining the databases that already exist and because of the equally great difficulty of reconciling databases that were created using incompatible data models or different data-collection procedures (Brackett 1994).

It will always be difficult to reconcile existing databases. But newly created databases may be another story. Whereas the old dystopian scenario focused its attention on a single centralized computer, a new scenario focuses on standardized data models -- data structures and identification and categorization schemes that emerge as standards, either globally or sectorally. Standards have contributed to the rise of institutions because they permit reliable knowledge of distant circumstances (Bowker 1994; Porter 1994). Many people once thought that it was impossible, for example, to categorize and intermingle grain from different farms (Cronon 1991). Once this was finally achieved, after several false starts, it became possible to trade commodities in Chicago and New York. Those cities became "centers of calculation" (Latour 1987) that gathered and used information from geographically dispersed sources. Information from multiple locations was commensurable, to within some controllable degree of uncertainty, because of the standardization of categories and measurement schemes. As a result, those with access to this combined information possessed a "view from nowhere" (Porter 1994) that permitted them to enter a tightly bound relationship with numerous places they had never visited.

Standardized data models may produce a similar effect, and for precisely the same reasons: standardized information is more valuable when it crosses organizational boundaries, and standardized real-world circumstances are more easily administered from a distance. A firm that plans to exchange information with other organizations can facilitate those transactions by adhering to standards, and a firm that expects to consolidate with other firms in its industry can increase its own value by standardizing its information assets.

It now becomes possible to employ distributed object database technology (Bertino and Ozsu 1994) to share information in an extremely flexible fashion. To be truly dystopian, let us imagine a single, globally distributed object database. Every entity instance in the world, such as a human being, would be represented by a single object, and all information about that person would be stored as attributes of this object. Individual organizations would still retain control over their proprietary information, but they would do so using cryptographic security mechanisms. At any given time, each organization would have access to a certain subspace of the whole vast data universe, and it would be able to search the visible portion of that data space with a single query. Organizations wishing to make certain parts of their database public (for example, firms wishing to make product information available to customers) would lift all security restrictions on the relevant attributes. Organizations with data-sharing agreements would selectively provide one another with keys to the relevant attributes in their respective segments of the data space. Real- time markets in data access would immediately arise. Lowered transaction costs would permit access to individual data items to be priced and sold. Data providers would compete on data quality, response time, and various contract terms (such as restrictions on reuse). Once the global database was well established, it would exhibit network externalities, and whole categories of organizations would benefit from joining it for the same reasons they benefit from joining interoperable networks such as the Internet. (Database researchers such as Sciore, Siegel, and Rosenthal (1994) speak of "semantic interoperability," but they are usually referring to systems that render existing heterogeneous databases interoperable by translating each database's terms in ways that render them commensurable with one another.)

Such a database would, of course, be extraordinarily difficult to get started. Bowker and Star (1994) illustrate some of the reasons in their study of an ambitious global project to standardize disease classifications. But the standardization of grain was difficult too. Standardization is a material process, not just a conceptual exercise, and standardized databases cannot emerge until the practices of identification and capture are standardized. The point is that considerable incentives exist to perform this work. On this view, privacy is not simply a matter of control over data; it also pertains to the regimentation of diverse aspects of everyday life through the sociotechnical mechanisms by which data are produced (Agre 1994).

5. Constructing Technology and Policy

Although powerful as stimuli to the imagination, the scenarios in the previous section are too coarse to account for the social contests through which privacy issues evolve. The authors in this volume paint a coherent and powerful picture of these contests -- a political economy of privacy that complements and extends the work of Gandy (1993). This section summarizes their contributions, focusing on the interactions among technology, economics, and policy. This theory starts with observations about the somewhat separable logics of technology, economics, and policy.

* Technological logic. Among the many factors influencing the evolution of technical systems, some are internal to the engineering disciplines that design them. Bijker (1987) refers to the "technological frames" that shape engineers' understandings of the problems they face and subsequently shape outsiders' understandings of the technologies themselves. Agre describes the metaphors -- processing and mirroring -- that help define a technological frame for the design of a computer system. Another aspect of this frame is the historically distant relationship between the designers of a system and the people whose lives the system's data structures represent. Although system designers' technological frame has changed slowly over the history of the computer, Bennett observes that the technology has improved rapidly in speed and connectivity. As a result, the underlying representational project of computing -- creating data structures that mirror the whole world -- has found ever-more-sophisticated means of expression in actual practices. Other technological systems -- for example, the infrastructures of transportation, communications, and finance -- embed their own disciplinary logic that shapes the privacy issues that arise within them.

* Economic logic. Privacy issues have evolved in the context of several trends in the global economy. Samarajiva points to the decline of the mass market and the proliferation of "compacks" -- packages of products and services that are, to one degree or another, adapted to the needs of increasingly segmented markets and even particular customers. The flexible production of compacks presupposes a decrease in the costs of coordinating dispersed manufacturing and service activities, and the marketing of compacks presupposes a decrease in the costs of tracking the market and maintaining tailored relationships with numerous customers. Samarajiva also points to the significance of network externalities (which strongly condition attempts to establish new information and communications networks) and the peculiar economics of information (which can be dramatically less expensive to distribute than to produce). These effects imply that classical economic models may be a poor guide to the success of new social institutions.

* Policy logic. The authors in this volume are particularly concerned with the dynamics of the policy process. Moving beyond a normative consideration of privacy policies, they reconstruct the evolution and the implementation of these policies. Their central claims have already been sketched. Bennett observes that privacy policy has emerged from a global network of scholars -- an "epistemic community" (Drake and Nicolaidis 1992) whose thinking develops in a coordinated fashion. Conflicts arise, as might be expected, through organizations' attempts to quiet privacy concern in the public sphere (Davies) and the countervailing initiatives of other policy communities (Phillips). Privacy issues are distinctive, though, in their rapid globalization and in the oft-remarked mismatch between their high level of abstract concern and their low level of concrete mobilization.

Expressed in this way, these points seem to hover outside history. Yet they are necessary preliminaries to consideration of the numerous modes of interaction among the respective logics. Perhaps the most striking aspect of this interaction is the recurring sense, remarked by Flaherty and Davies, that privacy is a residual category -- something left over after other issues have staked their claims. Privacy often emerges as a "barrier" to a powerful technical and economic logic, and privacy policy, in its actual implementation, often seems to ratify and rationalize new technical and institutional mechanisms rather than derailing them or significantly influencing their character. Part of the problem, of course, is a mismatch of political power. But it is also a problem of imagination: when privacy concerns arise in response to particular proposals, they can appear as manifestations of a perverse desire to resist trends that seem utterly inevitable to their proponents and to broader institutional constituencies.

With a view to understanding this phenomenon, let me sketch some of the reciprocal influences among the technological, economic, and policy logics of privacy.

Technological -> economic Technology influences economic institutions in several ways. Perhaps most obvious, technology largely determines the practicalities of production, and classical economics holds technical factors constant when exploring the conditions of market equilibrium. In a more subtle way, technology serves a rule-setting function. The computers and networks of a stock trading system, for example, embody and enforce the rules of trading. The rule-setting function is also familiar from the previous section's analysis of technology's role in setting ground rules for the negotiation of individual relationships. Finally, technology affects economic institutions through its effect on information costs and other transaction costs. As Casson points out, these effects can be far-reaching and qualitative and can provide a major impetus for the economic monitoring that raises privacy concerns.

Technological -> policy Technology influences policy formation, Phillips observes, through the obduracy of its artifacts: once implemented, they are hard to change. Technological practices are obdurate as well; PETs represent a rare occasion on which technical practitioners have revealed fault lines within an existing body of technical practices, so that formerly inevitable technical choices begin to admit alternatives. Technology thus influences policy in a second way: by determining the range of technological options available for addressing public concerns. Flaherty's tale of medical prescription systems represents a rare occasion on which a data protection commissioner has been able to affect the architecture of a proposed system, but at least the option was technically available. The dynamics of privacy policy, as Bennett points out, have also been shaped by the global nature of computer networking, which is even more indifferent to jurisdictional boundaries than the telephone system. Technology also influences policy, finally, by providing the infrastructure for contesting policy, as in the case of privacy activists' use of the Internet.

Economic -> technological Technology acquires its obduracy, in part, through the economic dynamics of standards. Once established in the market as the basis for compatibility among large numbers of otherwise uncoordinated actors, standards tend to be reproduced regardless of their consequences for concerns such as privacy, or even regardless of their efficiency in a changed technological environment (David 1985). The standards themselves, meanwhile, arise through practices of computer system design that have been shaped across generations through the demands of business applications. The full extent of this influence is not generally visible to the practitioners, who inherit a seemingly rational body of practices and who encounter "problems" in the bounded, conventionalized forms that these practices define. Yet the influence becomes visible suddenly as technical practices are moved from applications that structure relationships within the scope of a business to applications that structure relationships with other parties. A current example: Client-server systems arose on the premise that users do not wish do "see" the boundary between the client and the server. This is a reasonable commitment when that boundary has no legal or moral significance. It is not at all reasonable, however, for consumer applications -- such as commerce on the Internet -- in which that boundary corresponds precisely to the sphere over which individual users wish to maintain informational control. In such cases, it suddenly becomes important for the boundary between client and server to become visible, and the invisibility of this boundary becomes the raw material for rumors and scams.

Economic -> policy Flaherty notes several possible effects of economic phenomena on the policy process. Perceptions of the overall health of the economy may, other things being equal, influence the attention devoted to privacy issues. The organization of an industry affects its capacity to mobilize politically to define issues and shape regulatory regimes. Budgetary pressures on government agencies create organizational incentives for privacy-threatening initiatives that might not otherwise have materialized. Other economic phenomena affect the substance of policy issues and the institutional realities with which any policy will interact in its implementation. For example, Bennett points out that the economic properties of personal information have contributed to the global nature of privacy problems and to the global harmonization of privacy policy. Samarajiva's chapter is a concrete study in the complex arrangement of economic incentives that motivated one enterprise to establish relatively strong privacy policies. Because the system was only going to be viable if 80 percent of the potential members of the audience took specific actions to sign up, it became necessary to appease privacy fundamentalists by providing customers with a high degree of transparency and control over personal information. These policies were largely private rather than public, but they were congruent with the strong regulatory regime in Quebec. Similar considerations may explain the willingness of industries to submit to particular regulatory systems.

Policy -> economic Little work has been done to evaluate the economic impact of privacy policy. Reporting requirements impose costs, and rules about the secondary use of personal information affect business models and may sometimes determine their viability. On the other hand, when organizations review their information-handing practices with a view to implementing new privacy policies, they frequently discover opportunities for improvements in other areas. Privacy-protection measures may also reduce the economic costs associated with identity theft and other fraudulent uses of personal information. Policy may also influence economic institutions by shaping technology, though any massive effects of this type still lie in the future.

Policy -> technological Historically, privacy policy has not attempted to influence the basic architecture of information systems. Burkert and Agre, however, point to a joint report of Ontario's Information and Privacy Commissioner and the Netherlands' Registratiekamer (1995) that explores technological means for enhancing privacy. The privacy commissioners have focused on a subtle aspect of modern databases: although these databases have historically been designed on the assumption that data records can be traced back to the subjects they represent, such tracing is often unnecessary. Digital cash (Chaum 1992) is probably the best-known of the many schemes by which the tracing can be prevented or regulated to the subject's advantage by technical means while still providing organizations with the guarantees they need to conduct business. Even when the state does not mandate specific technical systems, Flaherty observes, data commissioners have an advocacy role of raising privacy issues early enough in the life cycle of an emerging infrastructure to raise consciousness about the implications of various technical options. At the same time, the US government has been attempting to regulate the spread of strong cryptography. Its sequence of "Clipper" proposals, described in part by Phillips, show the government learning how to intervene in the increasingly globalized dynamics of technical standard setting. If successful, these proposals would lead to a significant amount of new global infrastructure for the implementation of key escrow.

These considerations, drawn from several disciplines, begin to fill in the ambitious but oversimple picture that emerges from economic analysis alone. Privacy issues emerge as indicators of larger contests over the development of institutions, and the new landscape around technology and privacy emerges as a metaphorical Los Angeles: diverse, crowded, plagued by great inequalities and poor visibility, and subject to frequent earthquakes. Some hold that the Big One is upon us; others are more impressed with the damping effects of institutional inertia.

6. Policy Issues

The new landscape obviously poses significant challenges to privacy policy. Perhaps the greatest of these challenges derives from the greater range of available policy instruments. These instruments include the following:

* Privacy rules administered by a data protection commissioner. How can privacy officials enforce codes of fair information practices in times of rapid technological change? A movement toward policy convergence -- across borders, technologies, and categories of information -- would make privacy commissioners' work easier; increased fragmentation of policy or technology would make regulators' lives more complicated.

* Encouraging or requiring the use of privacy-enhancing technologies such as digital cash. Even though technological innovations hold considerable promise as a means of privacy protection, legislatures and privacy commissioners may have a limited ability to influence the fundamental architectures of technical systems. Nonetheless, they can raise consciousness, support research and standard-setting, participate in the development of demonstration systems, prepare model codes for those systems that do employ privacy-enhancing technologies, influence government systems procurement, and encourage academics to include privacy-enhancing technologies in textbooks and training curricula.

* Protocols for individualized negotiation of personal data handling. As opportunities emerge for individuals to customize privacy preferences, research should be conducted to evaluate alternative arrangements. These evaluations should employ a broad range of criteria, including ease of understanding, adequacy of notification, compliance with standards, contractual fairness and enforceability, appropriate choice of defaults, efficiency relative to the potential benefits, and integration with other means of privacy protection. Particular attention should be paid to uniformity of protocols across different industries and applications, so that consumers are not overwhelmed by a pointless diversity of interfaces and contracts.

* Outlawing certain technologies altogether, or limiting their use. To a certain extent, technical architectures encourage particular forms of interaction and particular types of relationship. Just as certain types of contracts are considered inherently unconscionable, it is possible that particular technologies may be found intrinsically unfair in their practical effects.

* Sectoral codes of practice. Sectoral codes were once viewed as alternatives to regulation. Now they are increasingly viewed as complements to an established legal framework. Sectoral codes provide opportunities to tailor codes of fair information practice to particular services. Nonetheless, it remains to be seen whether sectoral approaches to privacy protection will survive the rapid convergence of information flows in global computer networks.

* Establishing a broader legal basis for torts of privacy invasion. Although a vast amount has been written about the philosophical basis for various torts of privacy invasion (e.g., Schoeman 1984), these torts have remained narrow in scope and clearly inadequate to the technical and economic realities of the modern trade in personal information. As controversies arise in the new environment, common law countries may begin to explore new causes of action that formalize popular intuitions about the moral aspects of information exchange.

* Privacy-protection standards. Bennett describes an effort by the Canadian Standards Association to develop an auditable set of privacy standards along similar lines to the ISO 9000 standards. Although developed and administered by independent organizations, these standards can be referenced in legislation (for example as a condition of government contracts).

* "Bottom-up" education and training among system developers, employees with data handling responsibilities, and the general public. No matter how well-crafted a privacy code might be, privacy will only be protected if the necessary information practices are actually followed. Policy-makers need to understand how privacy issues actually arise in the daily activities of information workers, and organizational cultures need to incorporate practicable norms of privacy protection. Once established, these norms will only be sustained if the general public understands the issues well enough to make informed choices and to assert their rights when necessary.

These technical and policy solutions are potentially complementary; they comprise what Cavoukian and Tapscott (1997: 197-198) have called a "mosaic of solutions." It will require great effort to determine the appropriate combination of means to protect privacy in particular settings. PETs in particular must travel a long road from theory to practice, and it will be important to document and analyze their first applications.

The new technologies' great promise will also require theoretical innovation. As relationships are mediated by technology in more sophisticated ways, designers and policymakers will need more complex theories of agency and trust in technological environments. Perhaps most important, future research should clarify the relationship among technology, privacy, and association. Technologies for tracking people and conducting surveillance of public space risk chilling the freedom of association on which any possibility of democratic community is based. Yet the nature of that risk is still obscure. So long as privacy issues are located solely in the lives and relationships of individuals, it will be impossible to conceptualize either the potentials or the dangers of new media technologies for a democratic society.

* Acknowledgements

I greatly appreciate the contributions of Marc Rotenberg, as well as the comments and suggestions of Colin Bennett, Trotter Hardy, and Robert Horwitz.

* References

Agre, Philip E. 1994. Surveillance and capture: Two models of privacy. The Information Society 10, no. 2: 101-127.

Agre, Philip E. 1995. Reasoning about the future: The technology and institutions of Intelligent Transportation Systems. Santa Clara Computer and High Technology Law Journal 11, no. 1: 129-136.

Allen, Anita L. 1995. Uneasy Access: Privacy for Women in a Free Society. Rowman and Littlefield.

Bennett, Colin. 1992. Regulating Privacy: Data Protection and Public Policy in Europe and the United States. Cornell University Press.

Bertino, Elisa, and M. Tamer Ozsu, eds. 1994. Distributed and Parallel Database Object Management. Kluwer.

Bijker, Wiebe E. 1987. The social construction of Bakelite: Toward a theory of invention. In The Social Construction of Technological Systems, ed. W. Bijker et al. MIT Press.

Bowker, Geoffrey. 1994. Information mythology: The world of/as information. In Information Acumen: The Understanding and Use of Knowledge in Modern Business, ed. L. Bud-Frierman. Routledge.

Bowker, Geoffrey, and Susan Leigh Star. 1994. Knowledge and infrastructure in international information management: Problems of classification and coding. In Information Acumen: The Understanding and Use of Knowledge in Modern Business, ed. L. Bud-Frierman. Routledge.

Brackett, Michael H. 1994. Data Sharing: Using a Common Data Architecture. Wiley.

Branscomb, Lewis M., and James H. Keller, eds. 1996. Converging Infrastructures: Intelligent Transportation and the National Information Infrastructure. MIT Press.

Brenton, Myron. 1964. The Privacy Invaders. Coward-McCann.

Burnham, David. 1983. The Rise of the Computer State. Random House.

Casson, Mark. 1994. Economic perspectives on business information. In Information Acumen: The Understanding and Use of Knowledge in Modern Business, ed. L. Bud-Frierman. Routledge.

Cavoukian, Ann, and Don Tapscott. 1997. Who Knows: Safeguarding Your Privacy in a Networked World. McGraw-Hill.

Chaum, David. 1992. Achieving electronic privacy. Scientific American 267, no. 2: 96-101.

Clarke, Roger. 1994. The digital persona and its application to data surveillance. Information Society 10, no. 2: 77-92.

Coase, Ronald H. 1937. The nature of the firm. Economica NS 4: 385-405.

Cronon, William. 1991. Nature's Metropolis: Chicago and the Great West. Norton.

David, Paul A. 1985. Clio and the economics of QWERTY. American Economic Review 72, no. 2: 332-337.

Drake, William J., and Kalypso Nicolaidis. 1992. Ideas, interests, and institutionalization: `Trade in services' and the Uruguay Round. International Organization 46, no. 1: 37-100.

Ernst, Morris L., and Alan U. Schwartz. 1962. Privacy: The Right to Be Let Alone. Macmillan.

Flaherty, David H. 1989. Protecting Privacy in Surveillance Societies: The Federal Republic of Germany, Sweden, France, Canada, and the United States. University of North Carolina Press.

Fried, Charles. 1968. Privacy (A moral analysis). Yale Law Journal 77, no. 1: 475-493.

Gandy, Oscar H. Jr. 1993. The Panoptic Sort: A Political Economy of Personal Information. Westview.

Goffman, Erving. 1957. The Presentation of Self in Everyday Life. Anchor.

Gostin, Lawrence O. 1995. Genetic privacy. Journal of Law, Medicine, and Ethics 23, no. 4: 320-330.

Information and Privacy Commissioner (Ontario) and Registratiekamer (The Netherlands). 1995. Privacy-Enhancing Technologies: The Path to Anonymity (two volumes). Information and Privacy Commissioner (Toronto) and Registratiekamer (Rijswijk).

Latour, Bruno. 1987. Science in Action: How to Follow Scientists and Engineers Through Society. Harvard University Press.

Lyon, David, and Elia Zureik, eds. 1996. Computers, Surveillance, and Privacy. University of Minnesota Press.

Nugter, A. C. M. 1990. Transborder Flow of Personal Data within the EC: A Comparative Analysis of the Privacy Statutes of the Federal Republic of Germany, France, the United Kingdom, and the Netherlands. Kluwer.

Packard, Vance. 1964. The Naked Society. McKay.

Porter, Theodore M. 1994. Information, power and the view from nowhere. In Information Acumen: The Understanding and Use of Knowledge in Modern Business, ed. L. Bud-Frierman. Routledge.

Regan, Priscilla M. 1995. Legislating Privacy: Technology, Social Values, and Public Policy. University of North Carolina Press.

Registratiekamer. See Information and Privacy Commissioner.

Rotenberg, Marc. 1996. Personal communication.

Schoeman, Ferdinand David, ed. 1984. Philosophical Dimensions of Privacy: An Anthology. Cambridge University Press.

Sciore, Edward, Michael Siegel, and Arnon Rosenthal. 1994. Using semantic values to facilitate interoperability among heterogeneous information systems. ACM Transactions on Database Systems 19, no. 2: 254-290.

Smith, H. Jeff. 1994. Managing Privacy: Information Technology and Corporate America. University of North Carolina Press.

Smith, Robert Ellis. 1979. Privacy: How to Protect What's Left of It. Anchor Press.

Turkington, Richard C. 1990. Legacy of the Warren and Brandeis article: The emerging unencumbered Constitutional right to informational privacy. Northern Illinois University Law Review 10: 479-520.

Westin, Alan F. 1967. Privacy and Freedom. Atheneum.

Williamson, Oliver E. 1975. Markets and Hierarchies: Analysis and Antitrust Implications. Free Press.

(c) Copyright 1997 by Philip E. Agre. All rights reserved.